A method for ground scene matching navigation and positioning of unmanned aerial vehicles based on remote sensing images

Abstract

The invention provides a method for ground scene matching, navigation and positioning of an unmanned aerial vehicle (UAV) based on remote sensing images, which is characterized in that the method comprises the following steps: step 1: acquiring a remote sensing image with detected feature vectors from a reference library in a satellite cruise system; step 2: correcting a real-time image taken when the UAV performs a task by using the remote sensing image to generate a registration image; step 3: extracting features from the registration image by using a SIFT algorithm to generate a first feature image; step 4: matching feature vectors of the first feature image with the remote sensing image by using a quill group algorithm to generate a second feature image; step 5: completing the positioning of the UAV by solving the second feature image to obtain three-dimensional data of the UAV at the moment of aerial photography; the invention solves the problem of ground scene matching, navigation and positioning of UAVs, and can provide guidance for the development and construction of matching, navigation and positioning technology in the aerospace field.

Description

Technical Field

[0001] The present invention relates to the field of positioning and navigation technology, and in particular to a ground scene matching navigation and positioning method for an unmanned aerial vehicle based on remote sensing images. Background Art

[0002] Target positioning is a special and critical task in the aerospace field, and it is also a research hotspot in the aerospace field in recent years, especially the study of how to make full use of satellite remote sensing image data to accurately analyze and locate targets from space. Indeed, satellite remote sensing maps are one of the key technologies in my country's aerospace and national defense security fields. In the civilian field, although buildings, mountains, streets, garden landscapes and other scenes can be clearly seen, they cannot meet application requirements in terms of accurate navigation and positioning, close-range image tracking and recognition, and satellite remote sensing close-range multi-target tracking, becoming a problem that needs to be solved urgently in the aerospace field.

[0003] Unmanned aircraft, referred to as "drones", are unmanned aircraft controlled by radio remote control equipment and self-contained program control devices. They are small in size, low in cost and easy to use. In recent years, they have been highly praised by major countries in the world and have stepped up their research and development. With the diversification of modern aerospace environment and the further development of science and technology, ground scene matching navigation and positioning technology has been continuously innovated, especially the coordinated action of satellite remote sensing navigation (GPS) and drones. In recent years, it has become one of the most widely used matching navigation and positioning technology methods in the aerospace field. It has many advantages such as high accuracy, strong real-time performance and fast response speed, and has become a hot research topic in my country's aerospace field. However, high-altitude drones generally have the disadvantages of large positioning errors in complex terrain areas and lack of continuous positioning capabilities for maneuvering targets. In particular, the traditional high-altitude drone real-time positioning method has poor target calculation accuracy, and the accuracy of drones decreases as the flight distance increases. It faces huge challenges when actually cooperating with satellite remote sensing navigation technology. Therefore, studying the scene matching navigation and positioning technology of low-altitude unmanned reconnaissance aircraft + satellite remote sensing images is of great significance to accelerate the development of my country's aerospace technology. Summary of the invention

[0004] Aiming at the problem of ground scene matching navigation and positioning of UAVs, this paper studies the ground scene matching navigation and positioning technology of UAVs based on remote sensing images, focusing on the realization of fast image matching algorithms in embedded environments, and targeting scenes such as cities and forests at all times of the day. The advantages of UAV remote sensing and GPS navigation and positioning technologies are integrated and complemented, and based on prior information such as general satellite remote sensing images, combined with visible light landscape images taken by UAVs, efficient ground scene matching navigation and positioning of UAVs is achieved.

[0005] In order to solve the problems existing in the prior art, the present invention adopts the following technical solutions to implement:

[0006] A method for ground scene matching navigation and positioning of an unmanned aerial vehicle based on remote sensing images, the method comprising the following steps:

[0007] Step 1: Obtain a remote sensing image with detected feature vectors from a reference image library in a satellite cruise system;

[0008] Step 2: Use remote sensing images to correct the real-time images taken by the drone when performing tasks to generate registration images;

[0009] Step 3: Extract features from the registered image using the SIFT algorithm to generate a first feature image;

[0010] Step 4: performing feature vector matching between the first feature image and the remote sensing image by using a cystic swarm algorithm to generate a second feature image;

[0011] Step 5: The second feature image is solved to obtain the three-dimensional data of the drone at the moment of aerial photography to complete the positioning of the drone.

[0012] Furthermore, the second feature image generation process in step 4 is as follows:

[0013] Step 101: Initialize the tunicate TSA according to the feature vectors of the first feature image and the remote sensing image, that is:

[0014] P p (x) = lb(x) + R(ub(x) - lb(x))

[0015] Among them: Each P p (x)(1≤p≤N) is the feature vector of a tunicate individual, i.e., the feature vector in the first feature image, N is the population size of the tunicate, lb(x) and ub(x) are P respectively. p (x) is the lower and upper bounds, R is a random number between [0,1];

[0016] At the same time, the parameters of the encapsulated swarm algorithm are initialized, and the initial value T of the number of iterations of the encapsulated swarm algorithm is set to 1.

[0017] Step 102: Match the first image feature vector with the remote sensing image feature vector by the following formula to generate a matching value F(P p (x)

[0018]

[0019] Among them, F(P p (x)) is the matching value, P p (x) i is the i-th dimension in the feature vector of the first feature image, y iis the i-th dimension of the remote sensing image feature vector;

[0020] Step 103: Iteratively update the tunicate TSA by introducing a new operator of the average individual of the tunicate population in the following way

[0021]

[0022] in, is the eigenvector of the first eigenimage of the tunicate group, r1, r2, r3, r4, r5, r6 are all random numbers uniformly distributed between [0, 1], To avoid The position vector of the conflict, is the feature vector of the real-time image that is most similar to the registered image;

[0023] Furthermore, the new operator of the average individual of the tunicate population is introduced Generated by the following formula:

[0024]

[0025] is the average position of the tunicate population, N is the population size of the tunicate population, r and is a random number uniformly distributed between [0,1];

[0026] The specific group behavior of the tunicate population is p (x) While implementing iterative updates, it is also necessary to The iterative update method is as follows:

[0027]

[0028] in, Indicates P p (x) is the updated position of the next generation of the swarm individual, c2 is a random number between [0,1].

[0029] Step 104: The number of iterations of the capsule group is T=T+1, and it is determined whether the number of iterations T reaches the preset maximum number of iterations T max , where T max is the set value, generally 100; if T≤T max , return to step 102 for fitness calculation, and repeat steps 102 and 103; conversely, if T>T max , then the algorithm process ends, at this time That is, the real-time image feature vector that matches the reference image feature vector y.

[0030] Beneficial Effects

[0031] The present invention has the following beneficial effects:

[0032] (1) Aiming at the problem of ground scene matching navigation and positioning of UAVs, this paper studies the ground scene matching navigation and positioning technology of UAVs based on remote sensing images, focusing on the implementation of fast image matching algorithms in embedded environments, and targeting scenes such as cities and forests at all times of the day. The present invention integrates and complements the advantages of UAV remote sensing and GPS navigation and positioning technologies, and based on prior information such as general satellite remote sensing images and combined with visible light landscape images taken by UAVs, can achieve efficient ground scene matching navigation and positioning of UAVs.

[0033] (2) P in the traditional quilt group algorithm p The iterative update of (x) is affected by The influence is large, resulting in low matching accuracy of feature vectors between the reference image and the real-time image. The advantage of the improved quilt group algorithm in the present invention is that it introduces the average individual of the quilt group population and proposes a new operator based on the average individual of the quilt group population. Avoid the problem that the algorithm optimization is only affected by the population of tunicate The guidance improves the matching accuracy of each feature point in the reference image and the feature vector of the real-time image of the UAV. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 The present invention is a flow chart of a method for ground scene matching navigation and positioning of an unmanned aerial vehicle based on remote sensing images. DETAILED DESCRIPTION

[0035] The following is combined with Figure 1 The present invention is described in detail:

[0036] The applicant has made a detailed illustration and description of the embodiments of the present invention in conjunction with the drawings in the specification. However, this is for the purpose of helping readers better understand the spirit of the present invention rather than limiting the scope of protection of the present invention. On the contrary, any improvement or modification based on the spirit of the present invention should fall within the scope of protection of the present invention.

[0037] In order to solve the problems existing in the prior art, the present invention adopts the following technical solutions to implement:

[0038] A method for ground scene matching navigation and positioning of an unmanned aerial vehicle based on remote sensing images, the method comprising the following steps:

[0039] Step 1: Obtain a remote sensing image with detected feature vectors from a reference image library in a satellite cruise system;

[0040] Step 2: Use remote sensing images to correct the real-time images taken by the drone when performing tasks to generate registration images;

[0041] Step 3: Extract features from the registered image using the SIFT algorithm to generate a first feature image;

[0042] Step 4: performing feature vector matching between the first feature image and the remote sensing image by using a cystic swarm algorithm to generate a second feature image;

[0043] Step 5: The second feature image is solved to obtain the three-dimensional data of the drone at the moment of aerial photography to complete the positioning of the drone.

[0044] Furthermore, the second feature image generation process in step 4 is as follows:

[0045] Step 101: Initialize the tunicate TSA according to the feature vectors of the first feature image and the remote sensing image, that is:

[0046] P p (x) = lb(x) + R(ub(x) - lb(x))

[0047] Among them: Each P p (x)(1≤p≤N) is the feature vector of a tunicate individual, i.e., the feature vector in the first feature image, N is the population size of the tunicate, lb(x) and ub(x) are P respectively. p (x) is the lower and upper bounds, R is a random number between [0,1];

[0048] At the same time, the parameters of the encapsulated swarm algorithm are initialized, and the initial value T of the number of iterations of the encapsulated swarm algorithm is set to 1.

[0049] Step 102: Match the first image feature vector with the remote sensing image feature vector by the following formula to generate a matching value F(P p (x)

[0050]

[0051] Among them, F(P p (x)) is the matching value, P p (x) i is the i-th dimension in the feature vector of the first feature image, y i is the i-th dimension of the remote sensing image feature vector;

[0052] Step 103: Iteratively update the tunicate TSA by introducing a new operator of the average individual of the tunicate population in the following way

[0053]

[0054] in, is the eigenvector of the first eigenimage of the tunicate group, r1, r2, r3, r4, r5, r6 are all random numbers uniformly distributed between [0, 1], To avoid The position vector of the conflict, is the feature vector of the real-time image that is most similar to the registered image;

[0055] Furthermore, the new operator of the average individual of the tunicate population is introduced Generated by the following formula:

[0056]

[0057]

[0058] is the average position of the tunicate population, N is the population size of the tunicate population, r and is a random number uniformly distributed between [0,1];

[0059] The specific group behavior of the tunicate population is p (x) While implementing iterative updates, it is also necessary to The iterative update method is as follows:

[0060]

[0061] in, Indicates P p (x) is the updated position of the next generation of the swarm individual, c2 is a random number between [0,1].

[0062] Step 104: The capsule group iteration number T=T+1, determine whether the iteration number T reaches the preset maximum iteration number T max , where T max is the set value, generally 100; if T≤T max , return to step 102 for fitness calculation, and repeat steps 102 and 103; conversely, if T>T max , then the algorithm process ends, at this time That is, the real-time image feature vector that matches the reference image feature vector y.

Claims

1. A method for ground scene matching navigation and positioning of unmanned aerial vehicles based on remote sensing images, characterized in that: The method comprises the following steps: Step 1: Obtain a remote sensing image with detected feature vectors from a reference image library in a satellite cruise system; Step 2: Use remote sensing images to correct the real-time images taken by the drone when performing tasks to generate registration images; Step 3: Extract features from the registered image using the SIFT algorithm to generate a first feature image; Step 4: Perform feature vector matching between the first feature image and the remote sensing image by using the cyst group algorithm to generate a second feature image; wherein: the second feature image generation process in step 4 is: Step 101: Initialize the tunicate TSA according to the feature vectors of the first feature image and the remote sensing image, that is: P p (x)=lb(x)+R(ub(x)-lb(x)) Among them: Each P p (x)(1≤p≤N) is the feature vector of a tunicate individual, i.e., the feature vector in the first feature image, N is the population size of the tunicate, lb(x) and ub(x) are P respectively. p (x) is the lower and upper bounds, R is a random number between [0,1]; At the same time, the parameters of the quilt group algorithm are initialized, and the initial value of the number of iterations of the quilt group algorithm is set to T = 1; Step 102: Match the first image feature vector with the remote sensing image feature vector by the following formula to generate a matching value F(P p (x) Among them, F(P p (x)) is the matching value, P p (x) i is the i-th dimension in the feature vector of the first feature image, y i is the i-th dimension of the remote sensing image feature vector; Step 103: Iteratively update the tunicate TSA by introducing a new operator of the average individual of the tunicate population in the following way in, is the eigenvector of the first eigenimage of the tunicate group, r1, r2, r3, r4 are all random numbers uniformly distributed between [0, 1]. To avoid The position vector of the conflict, is the feature vector of the real-time image that is most similar to the registered image; Step 104: The capsule group iteration number T=T+1, determine whether the iteration number T reaches the preset maximum iteration number T max , where T max As the set value, take 100; if T≤T max , return to step 102 for fitness calculation, and repeat steps 102 and 103; conversely, if T>T max , then the algorithm process ends, at this time That is, the real-time image feature vector that matches the reference image feature vector y; Step 5: The second feature image is solved to obtain the three-dimensional data of the drone at the moment of aerial photography to complete the positioning of the drone.

2. The method for ground scene matching navigation and positioning of unmanned aerial vehicles based on remote sensing images according to claim 1 is characterized in that: The new operator for averaging individuals in a tunicate population is introduced Generated by the following formula: is the average position of the tunicate population, N is the population size of the tunicate population, r and is a random number uniformly distributed between [0,1]; The specific group behavior of the tunicate population is p (x) While implementing iterative updates, it is also necessary to Iterative updates.

Images